quantlab.model.predefined.seed_ensemble

Seed ensembles: one model config trained under several random seeds.

SeedEnsemble(model, seeds) builds one member per seed, each a new instance of the model’s class on the model’s config with random_seed replaced. The members read the same data, so they share one collected panel and one feature request per prediction window; their predictions are averaged after a per-bar cross-sectional z-score (average_predictions). The ensemble satisfies the backtester’s Predictor protocol, so it is backtested, loaded and rebuilt from a run’s config.json like a single model.

Classes

SeedEnsemble	An ensemble of one model trained under several random seeds.
SeedEnsembleConfig	What a SeedEnsemble is rebuilt from: the wrapped model and the seeds.